CDOIQ Symposium 2024 | Tom Godden

News
More from CDOIQ Symposium 2024

Tom Godden

Enterprise Strategist and CxO Advisor

AWS

The role of governance in strategic data management and AI integration

Strategic data management is becoming increasingly critical as organizations navigate the complexities of today’s technological advancements.With the integration of AI technologies emerging as a pivotal factor for success, businesses are urged to adopt a strategic approach that focuses on solving specific problems with targeted data. Emphasizing incremental improvements and strong governance, this approach ensures continuous value delivery, according to Tom Godden (pictured), enterprise strategist and CxO advisor at Amazon Web Services Inc

Info
Transcript

Tom Godden | AWS

Tom Godden

Enterprise Strategist and CxO Advisor AWS

Tom Godden, an Enterprise Strategist at Amazon Web Services, emphasizes the importance of focusing on business value when dealing with data. He advises identifying specific business problems, gathering necessary data, and centralizing data into a common platform. Tom discusses the roles of Chief Data Officer and Chief Artificial Intelligence Officer in organizations, highlighting the need to define their responsibilities clearly. He suggests that organizations focus on enabling data access and adopt a proactive approach to data management. Tom emphasizes star... Read more

Tom Godden | AWS

>> Welcome back to Cambridge, Massachusetts, everybody. This is Dave Volante. We're here at the CDOIQ at the Hyatt Regency. Sanjeev Mohan is my co-host, and we're really pleased to have Tom Godden here. He is an Enterprise Strategist and CXO Advisor at Amazon Web Services. Tom, good to see you. Thanks for coming on.

>> Thank you very much for the opportunity. Great event. Excited to have a conversation today.

>> Yeah. In our hometown, you came in from the Coast.

>> Yeah, that's right.

>> But for our beautiful weather here, you're used to it where you live. We get it two months out of the year. So interesting background, Tom. We were talking before on your CIO, CXO background, but what does an enterprise strategist... Describe that sort of group inside of Amazon and what you guys do?

>> Yeah, it's a fantastic group. So I'm part of a group of 15 former Chief Information Officers, Chief Technology Officers from companies like Coca-Cola, McDonald's, Capital One, NASA. I was the Chief Information Officer of Foundation Medicine, just down the road here in Cambridge, Massachusetts. And as a team, we spend time having conversations with strategic customers around people, process, organization culture, how do they get their organizations right? And we do that through presentations. Did a great presentation today here at the event, but also in just direct conversations. And we have thousands of those per year, and we learn a lot and try to share that information with our customers.

>> So you guys like to work backwards, so let's start with the customer. My understanding is you helped Foundation get to the cloud and many in healthcare didn't want to do that, right? That was a scary proposition. And so what are customers telling you today about data? Obviously we hear all the time, "Our data is siloed, our data, it's really hard to make coherent." What are they telling you?

>> Well, that they want to get value out of their data. I think we went through the struggles in the early knots with people going, "Ah, data seems like a good idea, but it's expensive and I'll do it later." Now everyone is rushing to get to that data, but part of the struggle you mentioned is the data is siloed. The data may not be as well organized and managed and controlled. And so where I see organizations struggling is this desire to say, "Okay, we got to fix all of our data." You don't. You need to find a business problem. Identify just the data you need for that business problem and get that data right. Solve that business problem, lather, rinse and repeat. So the problem that I see lots of them doing is saying, "Oh, I need to pause, find all my data, gather all my data, catalog all my data, organize all my data." It sounds like a great way to spend $50 million and achieve nothing.

>> It's a very pragmatic approach.

>> Very AWS like.

>> Yes, what'd you expect? Right? But it's true because part of the thing is embrace the cloud, quick value, immediate value. If something doesn't work, turn it off, ramp it down, iterate, but have that eye towards continuous value delivery, not value someday, "When I get that big giant release done." How can we get value today out of this data? So that's where we spend a lot of time talking to organizations about how they do that.

>> So if I could carry that through, I'm inferring that the customers... You might say to the customer, "Look, the data silo is not your problem. It's understanding how to apply the data that matters for that business problem that you're trying to solve. Don't worry about the silo problem. Let's maybe solve that down the road. Get immediate value out of the data. Maybe you have a long-term strategy toward resolving and harmonizing some of the silo problem." Is that a right way to think about it? Or would you say, "No, we want to change the engine out while we're in midair?"

>> No, I think that you're generally right. I think we want to do all that with an eye towards how we can consolidate the data. So eventually, I think you're going to be better when you can bring that data together into a data lake, into a data platform to be able to analyze it, but you don't have to always make that move today in order to start getting value out of it. So have a vision towards centralizing your data into that common platform, but you don't necessarily have to do all of that today to get value.

>> But if you don't centralize it, which I agree, because that's a whole initiative by itself, and we've failed at it to have this enterprise data in-house-

>> Several times. Yeah.

>> Corporate data factory, we've been trying it for years, but then-

>> Maybe we just need a better name. We keep coming off with different names for it.

>> Correct. Yes. But then how do you reconcile these inconsistencies on the same data between different systems without even knowing those inconsistencies exist?

>> Well, it's a great question. And a strong governance program is part of that, automating the flow of your data so I have better control over my data, and beginning to establish that data lineage, that data versioning, that data library, that catalog that you have for data. And again, doing it iteratively, driven by business value. So do I need to solve the inconsistencies between this data and this data? Not necessarily, unless the use case I'm working on requires me to resolve that inconsistency, and then do it then to solve that business problem. What we did a lot, I led a lot of these, so I can say this, is we engineers looked at it and said, "Oh, those are inconsistencies. We should go fix them," to no business value. It just made us feel better that they were in all-

>> Yes.

>> Nice little rows and columns-

>> Job security.

>> And Sure, yeah. And I felt like I did something. At the end of the day, you're like, "Look, I-"

>> Yeah, busy work.

>> "I cleaned it all up. Isn't it all pretty?" You're like, "Well, what value did you get out of it?"

>> Right.

>> You're like, "Well, the value I got out of it is it's all clean and pretty." You're like, "No, you missed the point. It only needs to be as clean and pretty as it needs to create value." And we put the cart before the horse, and we did that as technologists. And so now what we're talking about is it's all driven by business value. Find that use case you need. Go get the data you need for it, master it, learn it, manage it, govern it. Just that data, deliver the value. Lather, rinse, repeat, do it over and over again.

>> I'd love to ask you a bunch of technical questions, but this is really a business conference. You're a former CIO, which to me, is a business role .

>> Which means I'm only good at Excel and PowerPoint at this point in my career.

>> There you go.

>> I'm an expert PowerPoint guy, if you need any.

>> We all started as developers.

>> Yeah, we did. We did.

>> We called ourselves programmers back then.

>> That's right. That's right.

>> Okay. So Andy Jassy famously said many times and many times in the queue, but we're comfortable being misunderstood for a while. And I think people generally have misunderstood your AI strategy, and frankly, some of that is your own fault. But regardless, we know that people look to .

>> We are playing a different game than other people.

>> That's fair. And we're a very heavy AWS user. We run our rag on your platform and run our business basically, with the exception of some Google Docs. But my question is working, again, backwards from the customer. Now, AI, most CIOs would tell you, "Look, don't worry about the technology. Worry about the people, the process and the technology will take care of it. It'll come and go."

>> Yep.

>> It's a little bit different this time because it's one of those technologies that everybody's talking about. It's on the morning shows, Good Morning America is talking about the AI, but still, it's yet another technology that you have to apply, to your point, for business outcomes. So how are people thinking about their AI strategies in relation to the Chief Data Officer role and this emerging role of a Chief AI Officer that we've been talking about a lot today? Do they go together? Is it peanut butter and jelly, or are there separate roles? What are you seeing?

>> Well, I think the risk that we have with the Chief Artificial Intelligence Officer is some of the same risk we've encountered with the Chief Data Officer, which is we expected them to work magic because we created the position. Why are they here? What are they responsible for? What outcomes are they trying to affect? How have they been enabled within the organization? Great. They report to the Chief Executive Officer, but it still isn't going to say that they have a remit to solve. We're expecting magic to happen. So I think what we need to make sure that we do is we define those is the why? What are they trying to affect? What is their responsibility that they have and have they been enabled to do that? And by and large, I'm seeing organizations continue to struggle with that. If I'm honest, I think they've struggled with that with the Chief Data Officer role as well. How does the Chief Data Officer role relate to the Chief Information Officer role? Now we're going to put a Chief Artificial Intelligence Officer role into this because we think it's going to help? No, this is more bureaucracy. It's not what we need. We need simplicity on this. We need clarity. And so I worry that as we create these roles that we're moving away from clarity, but it feels good because again, we did something. "I created a Chief Artificial Intelligence Officer, check."

>> But it's so interesting you said this because last night at dinner, Dave, this is what you and I were talking about. You start with why? What it's called is not that important.

>> Yes.

>> It's like, why are you doing this?

>> What's the outcome we're trying to affect?

>> What's the outcome we're trying to... And then how will we achieve that outcome? What you call it doesn't really matter as much.

>> And that's that Amazonian pragmatic, working backwards, identify the value you're trying to create-

>> Correct.

>> Obsess about just that and deliver it. And labels are meaningless.

>> I love this conversation because it is contrarian to a lot of the conversations we have-

>> Right.

>> But you know the saying, "The trend is your friend," but at the same time, if they follow lemmings, you might follow them right off at the cliff. My question relates to governance and this whole trend toward open table formats. It's like the big rage this year. Everybody wants to control their own data, "Don't give it to your . Don't give your data to a vendor, even us." Okay, that's strong messaging, but so how are customers thinking about that whole open data format, bringing any compute engine to the data and controlling my data? How does that relate to the pragmatic solving of a business problem? What would you say to customers that ?

>> I think they're struggling a little bit with it because if I want to keep all my data to myself and not share with anyone, well, that doesn't work if we all share the same opinion. Right? It only works if we agreed to do that, to share the data. And so I'm seeing the customers be more successful, that have a good governance program, that have good control over their data so they can enter into effective IP sharing types of agreements. But again, it comes back to you need a good governance program in place. It's something that's interesting, and hindsight's such a beautiful thing, where you look back and you go, "Well, of course." But when this whole generative AI thing started, you looked at the highly regulated industries and you said, "They're going to be laggards." Why? Because they're regulated. And we're like, "No. It turns out they're actually leading a lot of this charge." And the reason why they're leading this charge is because they had their house in order. They already needed to manage their data. They had good control over their data. They understood lineage and versioning and had data governance, and they had a lot of the things, not all, but they had a lot of the things that you needed where other ones that weren't as regulated had some, but not by and large. The regulated guys got a good leg up on it. And so I think they're leaning into these agreements on how can you share the data and be open. But one more thing I'll say on this is we're still struggling as organizations where our first instinct is, "How can I restrict this data?" As opposed to, "How can I make this data available for everyone?" And again, no one's saying be reckless and wasteful and share your private information, especially the regulated stuff. But the mindset needs to change to, "How do I enable people more through data?" As opposed to, "I'm going to restrict what people can do," and that starts with how you view data inside. I think that carries to how you view data outside.

>> So the original view of data governance was very defensive, and that's why no one wanted to do the job of data governance. You would never get promoted-

>> No.

>> For defending your data.

>> It was the office of no.

>> Yeah.

>> But they did all that boring work that you, Tom, was just talking about.

>> Correct. Yeah, yeah.

>> Yeah. But now that data governance is about enabling-

>> Exactly.

>> And exposing the data to bring some value, it's actually become a very good position to be in.

>> Well, and the ones that I see succeeding are the ones that are governing by enabling. So think of your favorite app store that you use, whatever app store you use on your phone. They actually have a lot of governance built into the app store. What you can install, how you can pay for it, the content that can be on there, but by and large, we accept that. And why do we accept that? Because it works, because I can install something on my phone in seconds. I know it's secure. I know when I install an update to my mail app, it's not going to crash my browser. It works. So that same mindset of, "They've made it easy for me to use and they've slipped the governance in underneath and I haven't seen it." I think when you can embody that inside of your organization and say, "Let me just make these data sets the easy data sets for you to use," they're easy to download, they're easy to access, they're easy to understand. "I built in all the governance behind them and the data application and all these other things," but you don't care because it's just so easy for you to get access to.

>> Correct.

>> It's the place you go for that data. And those that have embodied that, govern by enable, those are the ones that I'm seeing transform.

>> What about the folks who haven't? What's your message to them? Still, I think the data, it's declining. The percentage of customers we survey with our partner ETR, that are not pursuing GenAI, it's down under 20%, but it's still probably 15%, 16% that aren't. And when you ask them why, they say, "It's because it's too risky, it's moving too fast. Compliance, regulatory concerns, legal concerns." What do you tell those folks?

>> I think those are all valid concerns, and I think we've got to help them understand those concerns. A mistake I see organizations making is assuming the risk for all generative AI is extreme. And certainly, if I'm using generative AI to do tumor detection inside of X-rays, yeah, but if I'm using it to summarize meeting notes inside your organization, it's a completely different risk profile. So we need to help them understand the risk, build programs to mitigate the risk, but really help them have that appreciation. And part of it is because right now there's so much unknown, there's that fear. And so I think understandably, people have turned the risk meter up because they don't understand. And so candidly, that's part of my role, it's part of our message to AWS is to help them understand there is a path on this. There is ways to mitigate and to manage those things and find your way through it. The other message that I would give them is, "This afternoon would be a good place to start. Yesterday would've been better. It's time, you got to get going."

>> Because there is risk-reward too.

>> Yeah, for sure.

>> And so really getting started is-

>> But the good news is any seasoned executive has dealt with risk registers before.

>> Sure.

>> Right? And that's what we tell them. We're like, "Guys, okay, it's a different thing, but use the same principles that you did for anything else that you've managed risk and apply it here."

>> I'd love to get your reaction. As I was talking to... I guess it was an architect, CTO type at a big bank, New England Bank, huge bank. Everybody would know the name. And he said to me... You know the saying, "Go fast and break things?" He goes, "We have to now go fast, but don't break things."

>> Yeah.

>> That's different, Tom, than the traditional... Or maybe not, maybe not in financial services than the traditional risk.

>> No, you still don't want to break things. That doesn't mean that I can't iterate and experiment. I don't have to put it into production until it's worthy, but that doesn't mean that I need to sit by and just say, "I'm taking a pass on all this technology." Experiment, practice, get comfortable. You have risk requirements, you have regulatory requirements, meet them, but you can iterate and practice and you can learn. And again, summarizing your meeting notes is a way to start learning, helping your onboarding of your employees by providing them a conversation-

>> More risk.

>> Do you understand your benefits? Would you like an AI tool that can help decipher your benefits? So your wife goes, "Are we covered for this dental?" You're like, "I have absolutely no idea."

>> ""What's my number?"

>> I have dental insurance. That's as far as I could answer. Could we build a solution to help our employees do that?

>> Yeah. Taking a pass is a mistake.

>> Taking a pass is... It's the old Scott McNeely quote, right? "The worst thing you can do is nothing." Right?

>> It's a strategy, but not really.

>> It is. It's not a long-lasting one.

>> I talked to the CDO of this very large financial services company that has now a production product and that production... Asking questions, it's like, "Let's say I'm in Illinois and my family moves to Texas and I want to know what changes with my 529 or my IRA educational plan?" It's like, who would you call? Because it's very hard to get this information from a human being because the guy in Illinois would be like, "Well, I don't know anything about Texas. Let me talk to somebody in Texas." But now you can ask these questions.

>> And the beautiful thing, if you built a solution to do that, so risk is a lot lower. Still don't want to get it wrong-

>> Correct.

>> But my risk is lower. It can start getting people comfortable with it. It can start getting the creative juices flowing.

>> Yes.

>> It can start getting them to go, "Well, if I could build this to summarize my 529 plan eligibility, I wonder if we could build a solution for our contact center to do this?" And it can start getting that. This is a technology, and you mentioned it before, unlike I've seen a lot of others. Maybe the internet, it's similar, but it's different than the internet even that this one you learn by doing. It's not one that you can staff out and sit on the sidelines. The more you use it, the more you play with it, the more you understand it's boundaries, what it's capable of and what it's not capable of, and will help you as an executive then to guide that into solutions. So building these low risk ones, not only can you get value, remember that's what we were talking about, but it's also a great way to learn.

>> Yeah. Network effects in the internet were everything. Maybe not as applicable here, although scale is extremely important. But I want to come back to the value piece because the data, again, tells us that 40% of customers say they're stealing from other budgets to fund their GenAI. They're pushing out slightly, not huge ways, but we're definitely seeing a movement in the ROI, the payback period. More people are saying more than 12 months than those that you watch the All-In Pod and they say, "Oh, four guys that can build a billion-dollar company in a week."

>> Yes, .

>> "No problem." And people realize it's a little harder to get tangible value that's self-funding. You can hit a lot of singles. At what point do you feel like the GenAI initiatives need to become self-funding, get out of that experimentation phase in order to get what you guys like to call a flywheel effect? I don't think we're there yet. Maybe you would disagree, but at what point do we need to get there or are we there already?

>> I think that we're close. 2023 was about learning this technology. Exploded onto the stage, everyone's learning about it. 2024, people are trying to get going on it. It's more than proof of concept. It's trying to push things into production, but it's the time for value. It's time to get as you said, some singles, but get those singles to deliver value. But again, don't underestimate the value of lots of small, incremental, consistent improvements, as opposed to that big home run solution.

>> Did you see the Goldman Sachs report that came out last week that the economic benefit of GenAI is not there yet?

>> I haven't seen that one, unfortunately. No.

>> I see. Okay. It's getting a lot of publicity. 31 pages of pretty hard-hitting data that they've exposed, and it's their opinion. Their opinion is that GenAI projects are extremely expensive to deploy, and the returns that we are getting do not justify all that investment.

>> Well, I think again, got to do this for the right use case. I think there've been a lot of proof of concepts built as exercising the technology, not proving that they could have business value. And so those have underdelivered, right? We emphasize working backwards. That's where we started this conversation. Potentially, that means that you should use plain old, boring AI, not that cool generative AI. Start with the problem. Work backwards. Is it AI? Is it ml? Is it data analytics? Is it just a plain old software application? What is it that you need to have value? And maybe it's generative AI and the ones that we've seen succeed are the ones that are working backwards from that, not the ones that built a generative AI solution because they could. And they said, "Did that solve any problems?" And people are like, "No."

>> It's like you cannot do BI and you're doing AI. It's like first get your house in order. Get the BI to work.

>> A little bit. Yeah. And work backwards from the problem. Find the value. And if you can do that, don't underestimate the value of consistent incremental improvements that can be made.

>> All right, Tom, we got to leave it there. Thanks so much for your perspective, somewhat contrarian. We like it, we like that Amazon has a different perspective. Thanks for coming on the queue.

>> Thank you for having me.

>> You bet. All right, keep it right there. Sanjeev Mohan and Dave Volante from CDOIQ in Cambridge, Massachusetts. We'll be right back right after the short break.